LDP-noise-addition

Overview

This repository contains a Python script for adding Local Differential Privacy (LDP) noise to a dataset. The script demonstrates two types of noise addition methods: Laplacian noise and randomized response. It provides functions to add these noise types to both numerical and categorical columns of the dataset.

Script Details

• The script provides functions to add Laplacian noise and randomized response noise to numerical and categorical columns, respectively.
• It utilizes the numpy and pandas libraries for numerical operations and data manipulation.
• Make sure to specify the privacy parameter eps according to the desired level of privacy for the dataset.
• The noisy dataset is saved as a new CSV file with the suffix _eps_<epsilon_value>.csv in the Data/ directory.

Usage

1. Clone the repository to your local machine:

   git clone https://github.com/your_username/LDP-noise-addition.git
   

2. Install the required dependencies:

   pip install numpy pandas
   

3. Place your dataset in the Data/ directory. The script assumes the dataset is in CSV format.
4. Update the script (add_ldp_noise.py) with the correct path to your dataset and modify the categorical/numerical columns.
5. Run the script:

   python add_ldp_noise.py
   

6. The script will generate a new CSV file with the noisy dataset in the Data/ directory.

Dataset

The dataset used in this project is the "Adult Census Income" dataset, available from Kaggle. It contains various demographic and financial attributes of individuals.